DPC-VQA: Decoupling Quality Perception and Residual Calibration for Video Quality Assessment

2026-04-14

AI Summary

This work addresses the high cost of adapting existing multimodal large language model (MLLM)-based video quality assessment methods to new scenarios, which typically require extensive retraining and abundant subjective annotations. To overcome this limitation, the authors propose a novel decoupled paradigm that freezes the pretrained MLLM to provide a fixed perceptual quality prior and introduces only a lightweight calibration branch to learn residual corrections. This approach is the first to decompose video quality assessment into two stages—fixed perception and learnable calibration—dramatically reducing trainable parameters to less than 2% and annotation requirements to merely 20% of subjective labels. Despite this efficiency, the method achieves performance on par with state-of-the-art approaches on both user-generated content (UGC) and AI-generated content (AIGC) benchmarks, significantly enhancing deployment efficiency and generalization capability.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Multi-modal Vision
• Natural Language Processing: (Large) Language Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Recent multimodal large language models (MLLMs) have shown promising performance on video quality assessment (VQA) tasks. However, adapting them to new scenarios remains expensive due to large-scale retraining and costly mean opinion score (MOS) annotations. In this paper, we argue that a pretrained MLLM already provides a useful perceptual prior for VQA, and that the main challenge is to efficiently calibrate this prior to the target MOS space. Based on this insight, we propose DPC-VQA, a decoupling perception and calibration framework for video quality assessment. Specifically, DPC-VQA uses a frozen MLLM to provide a base quality estimate and perceptual prior, and employs a lightweight calibration branch to predict a residual correction for target-scenario adaptation. This design avoids costly end-to-end retraining while maintaining reliable performance with lower training and data costs. Extensive experiments on both user-generated content (UGC) and AI-generated content (AIGC) benchmarks show that DPC-VQA achieves competitive performance against representative baselines, while using less than 2% of the trainable parameters of conventional MLLM-based VQA methods and remaining effective with only 20\% of MOS labels. The code will be released upon publication.

Problem

• video quality assessment
• multimodal large language models
• mean opinion score
• model adaptation
• annotation efficiency

Innovation

• Decoupling
• Perceptual Prior
• Residual Calibration
• Video Quality Assessment
• Lightweight Adaptation